基于鱼眼镜头的运动目标跟踪  

Moving target tracking based on the fish-eye lens

作　　者：朱均超[1] 刘娜[1] 张宝峰[1] 

机构地区：[1]天津理工大学天津市复杂系统控制理论及应用重点实验室,天津300384

出　　处：《激光与红外》2013年第7期837-841,共5页Laser & Infrared

基　　金：国家自然基金项目(61172185);天津市高等学校科技发展基金项目(20100705)资助

摘　　要：针对运动目标在普通镜头下跟踪视角小,以及在复杂背景下容易丢失的问题,提出了一种基于鱼眼镜头的改进Mean-shift算法。首先通过SIFT算法提取运动目标初始帧,利用卡尔曼滤波算法预测下一帧运动目标位置,进而采用改进的Mean-shift算法进行运动目标的跟踪。实验结果表明利用鱼眼镜头配合本文改进的Mean-shift算法具有良好的跟踪效果,与传统的跟踪方法相比具有大广角、实时性、鲁棒性、准确性等特点。With normal lens for moving target tracking, the FOV is small and it is easy to lose the target under compli- cated background. To solve the problems, an improved mean-shift algorithm based on the fish-eye lens is put for- ward. Firstly the moving target＇ s initial frame is extracted by using SIFT algorithm,then the next frame with target＇ s location is forecasted by using the Kalman filtering algorithm, finally the object is tracked by using improved Mean- shift algorithm. The result proves that this improved mean-shift algorithm based on fish-eye lens has good tracking per- formance, compared with the traditional tracking method. It has the characteristics of wide FOV, real-time, robustness and good accuracy etc.

关 键 词：运动目标跟踪 SIFT算法 鱼眼镜头 卡尔曼 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]